Evolutionary algorithm for large margin nearest neighbour regression

Abstract

The concept of a large margin is central to support vector machines and it has recently been adapted and applied for nearest neighbour classification. In this paper, we suggest a modification of this method in order to be used for regression problems. The learning of a distance metric is performed by means of an evolutionary algorithm. Our technique allows the use of a set of prototypes with different distance metrics, which can increase the flexibility of the method especially for problems with a large number of instances. The proposed method is tested on a real world problem – the prediction of the corrosion resistance of some alloys containing titanium and molybdenum – and provides very good results.